- The Automated
- Posts
- OpenAI’s hidden role in FrontierMath exposed
OpenAI’s hidden role in FrontierMath exposed
Plus, unlock $100+ in rewards.

Hello and welcome to the Automated, your AI tour guide.
AI benchmarks have become crucial for assessing a model's ability and credibility.
But what happens when it’s revealed that an industry heavyweight had direct influence and first hand access to those benchmarks?
Here’s what we have for you today:
😲 The Controversy Over OpenAI's Influence on FrontierMath.
🥳 ChatGPT Now Lets You Customize Its Personality with Traits Like ‘Chatty’ and ‘Gen Z.’
📒 How to use ChatGPT in Apple Notes for macOS.
🎬 How to create realistic AI videos using Sora.
🤖 ChatGPT Prompt Of The Day: Mimic an Artist's Style.
😲 The Controversy Over OpenAI's Influence on FrontierMath.

Epoch AI, the nonprofit behind the math benchmark FrontierMath, is facing some serious heat for not disclosing that OpenAI was secretly backing the project until, well… pretty much the last minute.
In case you missed it:
FrontierMath, designed to test AI’s ability to tackle expert-level math problems, was a key part of OpenAI’s demo for its upcoming model, o3.
Now here’s the real kicker: Even many key contributors had no idea OpenAI was involved until it was made public.
And, as you can imagine, they’re not exactly thrilled. Some are even saying they might have passed on contributing if they’d known OpenAI would get exclusive access to the benchmark.
Six mathematicians who significantly contributed to the FrontierMath benchmark confirmed this is true - that they are unaware that OpenAI will have exclusive access to this benchmark (and others won’t). Most express they are not sure they would have contributed had they known.
— Carina Hong (@CarinaLHong)
11:45 PM • Jan 18, 2025
On social media, concerns are swirling that this lack of transparency could seriously damage FrontierMath’s credibility as an impartial test.
And it’s not just about the funding—apparently, OpenAI got an early look at many of the problems and solutions before anyone else even knew they existed. Awkward right?
Epoch AI’s associate director, Tamay Besiroglu, admitted they “made a mistake” and should have pushed harder for transparency from the start.
He explained that legal constraints limited what they could reveal, but in hindsight, they should have insisted on keeping contributors in the loop from day one.
To ease the tension, Besiroglu assured everyone that OpenAI has a “verbal agreement” with Epoch AI not to train its AI models on FrontierMath’s problem set benchmark—because, well, teaching to the test is a big no-no.
There’s also a secret “holdout set” designed to guarantee independent verification—which is essentially an AI pop quiz that OpenAI supposedly hasn’t seen.
But here’s where things get murky— Epoch AI’s lead mathematician, Elliot Glazer, admitted they haven’t actually verified OpenAI’s test results yet.
While he personally believes OpenAI played fair, he admitted they can’t officially confirm anything until further checks are done.
So where does that leave us?
This entire saga is a stark reminder of just how tricky it is to create trustworthy AI benchmarks—especially when industry heavyweights like OpenAI are involved.
It also highlights the critical importance of transparency, as without it, even the most well-designed math tests can start to look a little….. shady.
So, what’s your take? Should we trust that OpenAI didn’t cut any corners, or does the secrecy raise red flags for you?
Either way, we’re still waiting on OpenAI’s side of the story— let’s just hope they’ve got some answers soon. In the meantime, click here to dive deeper into the drama.
🎁 Referral Radar: Keep the Good Times Rolling!
Last time, our Atomic Prompts referral campaign was a hit—thanks to YOU!
The energy and engagement from our 16,000-strong community made it clear: sharing knowledge and unlocking rewards is a win-win.
So, we’re keeping the good times going with a brand-new campaign:
A book like Atomic Prompts helps you clarify ideas, share insights, and build lasting impact. It can open doors, spark conversations, and position you as a thought leader.
Plus, this referral campaign guides you step-by-step on The Automated’s journey with book creation with prompts.
Here’s how you get in:
✨Refer 1 Friend – $12 Value
Get the first 3 chapters of I, Leader—FREE! (I, Leader, retails on Amazon for $11.99—grab this leadership playbook and gain insights that are redefining how we lead and inspire.)
✨ Refer 3 Friends – $49 Value
Unlock exclusive Loom tutorial videos where I walk you through, step-by-step, how I used prompts to write and publish a book.
✨Refer 5 Friends – $100+ Value
Earn a 30-minute 1-on-1 class with me! We'll work together to craft your personalized strategy for using prompts to create your book.
📣 Let’s keep this momentum alive! Act now to claim these rewards and continue shaping the future with us.
🥳 ChatGPT Now Lets You Customize Its Personality with Traits Like ‘Chatty’ and ‘Gen Z.’

Get ready to connect with ChatGPT like never before!
OpenAI is revolutionizing how we interact with AI by introducing exciting new features that allow you to customize the chatbot’s personality.
Whether you want it to be “Chatty” or radiate a “Gen Z” vibe, these fresh customization options promise a more relatable and dynamic conversation experience.
Announced just last Friday, OpenAI’s revamped custom instructions menu empowers you to tweak ChatGPT’s tone, style, and personality to match your preferences seamlessly.
This update is rolling out on ChatGPT.com and the Windows desktop app, with plans to expand to mobile and MacOS desktop apps soon.
We've updated custom instructions to make it easier to customize how ChatGPT responds to you.
With the new UI, you can tell ChatGPT the traits you want it to have, how you want it to talk to you, and any rules you want it to follow.
— OpenAI (@OpenAI)
6:35 PM • Jan 17, 2025
However, users in the EU, Norway, Iceland, Liechtenstein, and Switzerland will have to hold tight for a little longer due to regional restrictions, as stated in the tweet on Friday.
But this update isn’t all about looks; It’s about making your interactions truly personal!
You can now add your preferred name and profession, enriching the chatbot’s responses.
Want ChatGPT to sound more “Encouraging” or reflect the fun flair of “Gen Z”? Simply pick the traits that resonate with you, and watch the conversation transform!
This personalization feature evolves OpenAI’s prompt engineering framework, making it simpler and more accessible for anyone to customize their AI buddy.
While it doesn’t introduce radical technical changes, it enhances the chatbot's usability, making it feel more like a companion than a tool.
It's worth noting that this new customization feature is distinct from ChatGPT’s memory function, which tracks user information for deeper understanding over time.
With over 300 million weekly users, OpenAI is dedicated to making ChatGPT even more engaging.
Recent enhancements like live web search and a “Canvas” interface for writing and coding projects reflect their commitment to enriching your user experience.
Are you ready to transform ChatGPT into your ideal conversational partner?
Click here to dive into the new customization features today.
🧱Around The AI Block
📒 How to use ChatGPT in Apple Notes for macOS.
🎬 How to create realistic AI videos using Sora.
🔎 Microsoft opens testing for Windows AI search.
🦾 OpenAI finalizes 'o3 mini' reasoning AI model version, to launch it soon.
🤑 Lawmakers press Meta, Apple, Google, and others on massive Trump donations.
👩⚖️ FTC says partnerships like Microsoft-OpenAI raise antitrust concerns.
🤯 OpenAI is trying to extend human life, with help from a longevity startup.
🛠️ Trending Tools
Symbolab: A powerful online math solver that can help you tackle any math problem, from basic arithmetic to advanced calculus.
Lemlist: Helps to build your lead list with verified emails, write and personalize at scale, and send cold emails that actually get customers.
Tidalflow: Helps to create, guide, and adapt workout plans to the rhythm of your life.
Phrasee: Uses AI to generate billions of marketing messages to your clients.
DeskSense: Helps in problem-solving, code writing, and content creation across emails, replies, summaries, articles, blog posts, and social media.
🤖ChatGPT Prompt Of The Day: Mimic an Artist's Style.
Imitating the style of a famous artist is a great way to master new techniques, refine your skills, and explore creative possibilities.
It’s a stepping stone to finding your unique voice as you naturally add your own flair. Plus, it’s a great way to stay inspired and overcome creative blocks!
Here's a prompt that can help you with that.
Create a Picasso-inspired portrait
Generate a Van Gogh-style landscape with bright colors and thick brushstrokes
Design a psychedelic artwork in the style of Salvador Dali
We've Compiled a List of Over 100 ChatGPT Power Prompts.
This should help streamline your interactions with ChatGPT and get the results you need more efficiently.
Best of all, It's free!

That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What do you think of today's email? |
Your feedback means a lot to us and helps improve the quality of our newsletter.
Reply